# Correlational Tractography

## Before practicum on Friday, please complete following:

### Review Paper

### Correlational tractography tutorial

## During practicum on Friday:

### Correlation tractography

Conventional fiber tracking | Differential fiber traacking | Correlational Tractography | |
---|---|---|---|

seed point | start at any white matter location | start at any white matter location | start at any white matter location |

propagation | propagate along fiber orientation | propagate along fiber orientation | propagate along fiber orientation |

termination criteria | anisotropy threshold, angular threshold | anisotropy threshold, angular threshold, threshold for the anisotropy decrease |
anisotropy threshold, angular threshold, threshold for the t-statistics |

### Connectometry

Connectometry: a statistical method using permutation test and bootstrap resampling to test the significance of correlational tractography.

### Steps

- Calculate t-statistics using Spearman rank-based correlation
- permuted versus non-permuted, respectively, both after bootstrap resampling

- Fiber tracking based on t-statistics
- Each fiber tracking from a seed is a statistical test.
- The test statisticsis the length of track.
- positive findings: length > L, negative finding: length < L

- FDR calculation at length L
- #tracts with length > L after permutation => #false positive
- #tracts with length > L without permutation => #false positive + #true positive
- FDR = (#false positive)/(#false positive + #true positive)

- Results
- Fixed L → report FDR
- FDR threshold → estimate L* → showing findings satisfying FDR threshold

### Types

- cross-sectional versus longitudinal study
- correlation with a categorical variable (e.g. control v.s. patient) or continuous variable (e.g. age)

study type | correlation type | examples of null hypothesis |
---|---|---|

cross-sectional | correlation with a categorical variable | the tractography with decreased metrics in group 1 is the same as those after group permutation. |

cross-sectional | correlation with a continuous variable | the tractography with decreased metrics correlated with age is the same as those after age permutation. |

longitudinal | increased of decreased | the tractography with decreased metrics is the same as those after random-permuting the sign of change. |

longitudinal | associated with a categorical variable | the tractography with decreased metrics in group 1 is the same as those after group permutation. |

longitudinal | associated with a continuous variable | the tractography with decreased metrics correlated with age is the same as those after age permutation. |

### Hands-on: cross-sectional analysis

data: SCA2 cross sectional database and demographics

- correlational tractography correlated with group
- report FDR given a length threshold
- assign FDR threshold and report findings
- high t-threshold (more localized) versus low t-threshold (less localized)

- correlational tractography correlated with group under partial correlation considering age/sex.
- stratified analysis using cohort selection (e.g. male and female).
- combined ROI/ROA/terminative (wk 3 course)
- post-hoc analysis
- identifying pathways using manual virtual dissection and recognition (wk 2 course)
- tract-metric scatter plot using tract-based analysis (week 5 hw).

### Hands-on: longitudinal analysis

data: SCA2 longitudinal database

- compute longitudinal change using [Step C2][Tools][Longitudinal Scans]
- use connectometry to answer the following question
- are there significant decrease of FA in the patient group?
- are there significant decrease of FA in the control group?
- are decreased FA in the control group significantly correlated with age?
- are decreased FA in the patient group significantly correlated with age?
- are there significantly more changes in the patient group than control group?
- are there significantly more changes in the patient group than control group in the brainstem?